Associating stereo images acquired with two cameras with different views for automatic stereoscopic digital image evaluation, e.g. for reconnaissance, robotics and virtual reality

摘要

The method comprises combining the intensity and characteristics of both images in a uniform manner, applying a smoothing function in only homogeneous image regions, parameterizing disparity functions using a Radial-Basis-Function network, using a linear model to compensate for intensity differences between corresponding points and using a gradient descent method to minimize the error magnitude during correspondence searching.